Key value scheduling method based on receiving end driving

Abstract

The invention discloses a key value scheduling method based on receiving end driving. When the key-value operation reaches the client, the key-value operation is not sent immediately, but waits in the client queue for the optimal scheduling opportunity, and furthermore, the client distributedly adjusts the threshold that it allows to send the key-value operation to control concurrency between the clients. According to the method, key value scheduling can be carried out according to the dynamic change performance of the server, when the performance of some servers is poor, key value operation can be pressed in a client queue to wait for the optimal scheduling opportunity, the processing speed of the server with the good performance is high, and therefore more key value operation can be distributed. Besides, the number of key value operations allowed to be sent by the clients is adjusted in a distributed manner, so that good cooperation between the clients is achieved, and the problem of sharp increase of tail delay caused by concurrency in a large-scale scene is effectively controlled. Compared with an existing self-adaptive scheduling method, the method has the advantage that the tail delay of key value operation is effectively reduced.

Description

technical field [0001] The invention relates to the field of distributed key-value databases, in particular to a key-value scheduling method driven by a receiving end. Background technique [0002] With the continuous development of Internet technology, a large amount of data is being continuously generated. Due to its good scalability and high availability, distributed databases have gradually become the core storage platform for many interactive applications. A terminal user request usually generates dozens or even hundreds of key-value operations. The tail delay of these key-value operations determines the response time of the entire terminal request, which in turn affects user experience and enterprise revenue. In order to improve the efficiency of parallel storage, each piece of data is usually backed up to multiple nodes in the cluster. At the same time, the load and performance of each server change dynamically. If the key is processed on a high-load or low-performanc...

AI Technical Summary

Problems solved by technology

[0005] Aiming at the current technical problem that ORA cannot achieve optimal scheduling when the key-value operation is waiting in the queue of the server, and the concurrency of the key-value operation is large, which leads to a sudden increase in the instantaneous load of the server, the tail delay of the key-value operation will be reduced. The present invention provides a A receiver-driven key-value in a key-value storage system that allows the tail latency of key-value operations to be neither affected by high server load or low performance, nor affected by high concurrency due to the increase in the number of clients scheduling method

Images